sclerosis subtypes
`
`Assessing brain atrophy rates in a large
`population of untreated multiple
`
`N. De Stefano, MD
`
`A. Giorgio, MD
`M. Battaglini, MSc
`M. Rovaris, MD
`MP. Sormani, PhD
`F. Barkhof, MD
`
`T. Korteweg, MD
`C. Enzinger, MD
`F. Fazekas, MD
`M. Calabrese, MD
`D. Dinacci, MD
`
`G. Tedeschi, MD
`A. Gass, MD
`
`X. Montalban, MD
`A. Rovira, MD
`
`A. Thompson, MD
`G. Comi, MD
`D.H. Miller, MD
`
`M. Filippi, MD
`
`Address correspondence and
`reprint requests to Prof. Nicola
`De Stefano, Quantitative
`Narroimaging Laboratory,
`qurtrnent of Neurological and
`Behavioral Sciences, University of
`Siena, Viale Bracci 2, 53100
`Siena, Italy.
`
`Supplemental data at
`www.neurology.org
`
`ABSTRACT
`
`Objective: To assess the time course of brain atrophy and the difference across clinical subtypes
`in multiple sclerosis (MS).
`
`Methods: The percent brain volume change (PBVC) was computed on existing longitudinal (2 time
`points) T1-weighted MRI from untreated (trial and nontrial) patients with MS. Patients (n = 963)
`were classified as clinically isolated syndromes suggestive of MS (CIS, 16%), relapsing-remitting
`(RR, 60%), secondary progressive (SP, 15%), and primary progressive (9%) MS. The median
`length of follow—up was 14 months (range 12- 68).
`
`Results: There was marked heterogeneity of the annualized PBVC (PBVC/y) across MS subtypes
`(p = 0.003), with higher PBVC/y in SP than in CIS (p = 0.003). However, this heterogeneity
`disappeared when data were corrected for the baseline normalized brain volume. When the MS
`population was divided into trial and nontrial subjects, the heterogeneity of PBVC/y across MS
`subtypes was present only in the second group, due to the higher PBVC/y values found in trial
`data in CIS (p = 0.01) and RR (p < 0.001). The estimation of the sample sizes required for demon-
`strating a reduction of brain atrophy in patients in a placebo-controlled trial showed that this was
`larger in patients with early MS than in those with the progressive forms of the disease.
`
`Conclusions: This first large study in untreated patients with multiple sclerosis (MS) with different
`disease subtypes shows that brain atrophy proceeds relentlessly throughout the course of MS,
`with a rate that seems largely independent of the MS subtype, when adjusting for baseline brain
`volume. Neurology® 2010;74:1868-1876
`
`GLOSSARY
`ANOVA = analysis of variance; CIS = clinically isolated syndrome; DMA = disease-modifying agent; EDSS = Expanded
`Disability Status Scale; FSL = FMRIB Software Library; Gd = gadolinium; MR = magnetic resonance; MS = multiple sclertr
`sis.- NBV = normalized brain volume; PBVC = percent brain volume change; PP = primary progressive; RR = relapsing-
`remitting; SP = secondary progressive.
`
`A number of MRI—based methods for computed estimation of brain volumesl'2 have prompted
`the use of brain atrophy as a measure of disease progression in multiple sclerosis (MS). How-
`ever, the interpretation of brain volume change measurements is not always straightforward in
`MS and a number of confounding factors such as disease stage and disease—modifying agents
`(DMA) need to be consideredz‘3
`
`The natural evolution of global brain volume changes at different MS stages and without the
`influence of DMA has been investigated for each subtype on patients belonging to the placebo
`
`From the Department ofNeurological and Bd'ravioural Sciences (N.D.S., A.G., M.B.), University ofSiena; Neuroirnaging Research Unit (M.R.,
`M.F.), Institute of Experimental Neurology, Division of Neuroscience, Scientific Institute and University Ospedale San Raifaele, Milan; Biostatistics
`Unit (M.P.S.), Department of Health Sciences, University of Genoa, Genoa, Italy; Department of Radiology (F.B., T.K.), VU University Medical
`Centre, Amsterdam, the Netherlands: Department of Neurology (C.E., F.F.), Medical University of Graz, Craz, Austria: The Multiple Sclerosis
`Centre ofVeneto Region (M.C.), First Neurology Clinic, Department of Neuroscienoes, University Hospital of Padua, Padua; Department of
`Neurological Sciences (D.D., C.T.), Second University of Naples and Institute Hermitage Capodimonte, Naples, Italy', Department of Neurology
`(All), University Hospital, Kantonsspital, Basel, Switzerland; Department of Neuroimmunoloy and Radioloy (XM., A.R.), Hospital Vall
`d'Hebron, Barcelona, Spain; Department of Brain Repair and Rehabilitation (A.T.), Institute of Neurology, University College London, London, UK;
`Department of Neurology (G.C.), Institute of Experimental Neurology, Division ofNeuroscience, Scientific Institute and University Ospedale San
`Rafiaele, Milan, Italy: and Department of Neuroinflarntnation (D.H.M.), Institute of Neurology, University College London, London, UK. M.R. 's
`currently affiliated with the Multiple Sclerosis Centre, Scientific Institute S. Maria Nasoente, Fondan'one Don Gnocchi, Milan, Italy.
`Studyjimdingr The Quantitative Neuroimaging Laboratory was supported in part by the Italian MS Foundation. The VU University Center for MS
`Research (OS—358(2) and T. Korteweg (00—425 MS) are supported by the Dutch MS Research Foun
`ion.
`,
`.
`rogen Exhlblt 2084
`Disclosure: Author disclosures are provided at the end of the article.
`
`Mylan v. Biogen
`Copyright © 2010 by AAN Enterprises, Inc.
`1868 Page 1 of 9
`Copyright © by AAN Enterprises. Inc. Unauthorized reproduction of this it}; Icfeoissplibhqgited.
`
`

`

`arms of clinical trials.4-21 Further, research
`studies have been performed on patient
`groups partially treated with DMA or with a
`limited number of subjects.22-26 To date, no
`study has assessed temporal brain volume
`changes in a large untreated MS population,
`directly comparing different subtypes.
`Thus, we collected a large number of exist-
`ing MRI data on untreated patients with dif-
`ferent MS subtypes and analyzed them with a
`fully automated method for the estimation of
`global brain volume changes.27 We aimed to
`assess 1) differences in annualized global brain
`volume changes; 2) potential differences in
`brain atrophy rates between patients from the
`placebo arms of clinical trials (trial data) and
`those who remained untreated for that given
`follow-up period (nontrial data); 3) the sam-
`ple sizes required to demonstrate a treatment-
`related reduction of brain atrophy progression
`in placebo-controlled MS trials.
`
`METHODS Study population. This is a European multi-
`center retrospective study based on the analysis of longitudinal
`magnetic resonance (MR) datasets (2 time points) of patients
`with different subtypes of MS who were either collected at differ-
`ent imaging laboratories while not taking any DMA or in the
`placebo arms of clinical trials. Minimum between-scan interval
`was 12 months. There was no limitation for the Expanded Dis-
`ability Status Scale (EDSS)28 score at study entry. The main in-
`clusion criterion was the complete absence of DMA use during
`the study period. This did not include the use of steroids, but all
`the patients with MS had to be corticosteroid-free for at least 1
`month before scanning.
`A total of 1,160 pairs of T1-weighted MRIs was collected
`from 193 patients with a clinically isolated syndrome (CIS) sug-
`gestive of MS, 642 with relapsing-remitting (RR), 192 with sec-
`ondary progressive (SP), and 133 with primary progressive (PP)
`MS.29 The MRIs were obtained from data of imaging laborato-
`ries (Amsterdam, Barcelona, Basel, London, Milan, Naples, and
`Padua) and placebo arms of clinical trials (ETOMS,30 CORAL,12
`European/Canadian Glatiramer Acetate Study,31 and ESIMS19).
`The median length of follow-up was 14 months (range 12–
`42), with the exception of 3 RR patients who had the second
`MRI scan at 48 months and 1 RR patient who had it at 68
`months. The median was 24 (range 12–30) for CIS, 14 (range
`12– 68) for RR, 24 (range 12– 40) for SP, and 13 (range 12– 42)
`for PP. Clinical (i.e., disease duration and EDSS score) and de-
`mographic (i.e., age and sex) information were collected.
`
`Standard protocol approvals, registrations, and patient
`consents. The study received approval from the local ethics
`committee and written informed consent was obtained from all
`study patients.
`
`MRI data and analysis. All MR scans were acquired at each
`center using for each patient the same MR procedure/sequences
`and scanner at both time points. None of the scanners were
`changed or underwent major upgrades during the study period.
`
`Conventional T1-weighted images were sent to the Quantitative
`Neuroimaging Laboratory of the University of Siena for central-
`ized analysis.
`Global brain volume changes over time were quantified us-
`ing the SIENA method,32 part of the FMRIB Software Library
`(FSL; www.fmrib.ox.ac.uk/fsl/). This registration-based method
`uses images from 2 time points to assess brain volume changes by
`estimating directly the local shifts in brain edges across the entire
`brain and then converting the edge displacement into a global
`estimate of percentage brain volume change (PBVC) between
`the 2 time points.
`An automated procedure of brain extraction able to im-
`prove removal of eyeballs and remaining nonbrain tissues33
`was implemented in SIENA for a more accurate estimation of
`brain atrophy.
`The scans were all T1-weighted pairs of images obtained
`either prior to or after injection of gadolinium (pre-Gd and post-
`Gd). Slice thicknesses (range 1.2–5 mm) were identical in each
`pair of images. It was shown that use of different image types
`(pre-Gd and post-Gd T1-weighted images)2,34 and slice thick-
`nesses32 does not systematically affect SIENA measurements.
`However, differences in both T1-weighted image type and slice
`thicknesses were corrected for during the analysis.
`
`Statistical analysis. Changes in EDSS score (⌬EDSS) and
`PBVC were annualized (i.e., ⌬EDSS/y and PBVC/y) to account
`for differences in the length of follow-up between the 2 scans.
`A univariate analysis of variance (ANOVA) followed by pair-
`wise post hoc comparisons were used to compare PBVC/y across
`the different MS subtypes. A multivariate ANOVA was used to
`adjust the comparisons for age, sex, disease duration, data source
`(trial/nontrial), T1-weighted image type (pre-Gd and post-Gd),
`and slice thickness. Furthermore, PBVC/y values were compared
`across the disease subtypes by correcting for the baseline normal-
`ized brain volume (NBV) as measured on the T1-weighted im-
`age by using the cross-sectional version of SIENA (SIENAX),
`also part of FSL.
`Correlations (unadjusted for baseline values) of PBVC/y
`with demographic (age and sex) and clinical (disease duration,
`EDSS score at baseline, ⌬EDSS/y) features were analyzed using
`the Spearman rank correlation coefficient.
`SPSS software v11.0 (SPSS Inc., Chicago, IL) was used to
`perform statistical calculations. A 2-tailed p value of 0.05 was
`used as the cutoff for significance.
`The sample size required to demonstrate a treatment-related
`reduction in brain atrophy progression in placebo-controlled MS
`trials was estimated using PBVC/y as the primary outcome for
`each disease subtype. This was estimated to have a power of 90%
`at a confidence level of 5% for each disease subtype and to detect
`a treatment effect of 30%, 50%, and 70%. A nonparametric
`approach based on Monte Carlo simulations was used. The sam-
`ple size was estimated by the nonparametric Mann-Whitney U
`test comparing PBVC/y between the 2 arms.
`
`RESULTS Clinical and demographic information of
`the study population as well as MRI features are
`summarized in table 1.
`Out of 1,160 pairs of T1-weighted images, 197
`were excluded from the analysis for unsatisfactory
`quality (n ⫽ 143) or incomplete demographic or
`clinical information (n ⫽ 54). The final number of
`study subjects was 963, consisting of CIS (n ⫽ 157,
`
`Neurology 74 June 8, 2010
`
`1869
`
`Page 2 of 9
`
`

`

`I Table 1
`
`Demographic, cllnlcal, andmagnetic resonancefeatures ofstudy patlents
`RR
`SP
`All patients
`CIS
`(n = 963)
`(n = 157, 16%)
`(n = 579, 60%)
`(n = 138, 15%)
`
`|
`
`PP
`(n = 88, 9%)
`
`Age,y, mean (SD)
`Sex, n (96)
`
`Males
`
`Females
`
`Disease duration, y, mean (SD)
`Baseline EDSS
`
`Median
`
`Range
`
`EDSS change/y
`Median
`
`Range
`
`37.88 (10)
`
`31.35 (8.11)
`
`36.18 (8.24)
`
`45.29 (8.58)
`
`48.93 (11)
`
`333 (35)
`
`630 (65)
`
`58 (37)
`
`99 (63)
`
`725 (7.49)
`
`0.64 (1.45)
`
`2
`
`0 to 8
`
`0
`
`1
`
`0 to 4.5
`
`0
`
`160 (28)
`
`419 (72)
`
`7 (6.45)
`
`2
`
`0 to 8
`
`0
`
`63 (45)
`
`76 (55)
`
`52 (59)
`
`36 (41)
`
`15.44 (8.20)
`
`7.77 (6.88)
`
`5
`
`5
`
`1.5 to B
`
`1.5to7.5
`
`O
`
`0.2
`
`—3to3
`
`—3t02
`
`—1.5t03
`
`—1t02.5
`
`—1.5t03
`
`T1-weighted image type. n(%)
`
`Post-Gd
`
`432 (45)
`
`18 (11)
`
`348 (60)
`
`44 (32)
`
`95 (68)
`
`22 (25)
`
`66 (75)
`
`Pro-Gd
`
`Source. n We)
`
`Nontrial
`
`Trial
`Slice thickness
`
`1.2 mm
`
`1.5 mm
`
`3 mm
`
`4 mm
`
`5 mm
`
`531 (55)
`
`139 (89)
`
`231 (40)
`
`390 (40)
`
`573 (60)
`
`64 [7)
`
`16(2)
`
`446 (46)
`
`54(5)
`
`383 (40)
`
`74 (47)
`
`83 (53)
`
`21 (14)
`
`NA
`
`13(8)
`
`NA
`
`123 (78)
`
`173 (30)
`
`406 (70)
`
`37 (6)
`
`NA
`
`416 (72)
`
`49 (8)
`
`77 (14)
`
`54 (39)
`
`85 (61)
`
`3 (2)
`
`NA
`
`NA
`
`5 (4)
`
`131 (94)
`
`100
`
`NA
`
`3 (4)
`
`16 (18)
`
`17 (19)
`
`NA
`
`52 (59)
`
`Abbreviations: CIS = clinically isolated syndrome; EDSS = Expanded Disability Status Scale; Gd = gadolinium; NA = not
`available; PP = primary progressive; RR = relapsing-remitting; SP = secondary progressive.
`
`16%), RR(n = 579, 60%), SP (11 = 139, 15%), and
`PP (n = 88, 9%).
`
`Comparisons across MS subtypes. All patients. There
`
`was heterogeneity of PBVC/y across MS subtypes
`(PBVC/y, mean t SD: CIS = —0.40% i 0.47%,
`RR = —0.49% i 0.65%, SP = —0.64% i 0.68%,
`
`PP = —O.S6% i 0.55%,p = 0.003), with the pair—
`wise comparisons showing higher PBVC/y in SP pa-
`tients than in patients with CIS (p = 0.003) (figure
`1A). The between-group diflerence was maintained
`after correcting for age, sex, disease duration, slice
`thickness, data source (trial/nontrial), and T1-
`
`weighted image type (pre—Gd and post—Gd) (p <
`0.001 at multivariate analysis).
`the 47 subjects
`Among the patients with CIS,
`who converted to clinically definite MS showed
`higher PBVC/y than the 110 subjects who did not
`(—0.51% i 0.48% vs —0.35% i‘ 0.47%, p =
`0.04).
`
`As expected, baseline NBV was different across
`MS subtypes (NBV, mean : SD: CIS = 1,169 i 47
`cm}, RR = 1,140 i 53 cm3, SP = 1,089 i 50 cm},
`
`PP = 1,097 i 51 cm3,p < 0.001), with all pairwise
`
`comparisons across MS subtypes showing difl‘crences
`(p < 0.001) (with the exception of the comparison
`between SP and PP) (figure 1A). Interestingly, when
`PBVC/y values were corrected for the baseline NBV,
`the heterogeneity of PBVC/y across MS subtypes dis—
`appeared (p = 0.90) (figure 1B). It should be noted
`
`that despite significant diflerences, there was a de-
`gree of overlap for baseline NBV across MS sub-
`types. Figure 2 illustrates this overlap for CIS and
`SP, which represent
`the 2 extreme situations.
`Thus, controlling for baseline NBV should not
`lead to major extrapolation in comparing PBVC/y
`values across MS subtypes.
`Trial vs nontrial data. The heterogeneity of
`PBVC/y across MS subtypes detected on the whole
`population was still present when the analysis was
`performed on nontrial data (PBVC/y, mean i SD:
`CIS = —0.29% i 0.43%, RR = —0.34% i
`045%, SP = —0.65% i 0.65%, PP = —0.56% 1‘
`
`0.55%, p < 0.001). However, when the analysis was
`selectively performed on trial data, there was no het-
`erogeneity of PBVC/y across MS subtypes (PBVC/y,
`mean i SD: CIS = —0.49% i 0.50%, RR =
`
`1870 Page 3 of 9
`Copyright © by AAN Enterprises. Inc. Unauthorized reproduction of this article is prohibited.
`
`Neurology 74 June 8, 2010
`
`

`

`I Figure 1
`
`Brain atrophy measures In the different multlple selerosls(MS) subtypes
`
`l
`
`A 1.5
`1.2
`
`NBV
`
`PBVCIy
`
`m
`
`
`
`
`
`PBVCIycorrectedforbaselineNBV
`
`
`
`0.9
`
`0.6
`
`0.3
`
`0.0
`
`-0.3
`
`-0.6
`
`-0.9
`
`-1.2
`
`-1.5
`
`0.4
`
`0.2
`
`0.0
`
`.02
`
`-o.4
`
`-0.6
`
`-o.a
`
`-1.0
`
`-1.2
`
`-1.4
`
`(A)Values of percent brain volume change (PBVC)Iy (blue) and normalized brain volume (NBV) (expressed in liters, red) in the
`different MS subtypes. (B) Values of PBVC/y corrected for baseline NBV. Columns and error bars represent means and 80s
`of atrophy measures. Note the similar PBVCIy in the different MS subtypes when data are corrected for baseline NBV. See
`Results for details. CIS = clinically isolated syndrome; PP = primary progressive; RR = relapsing-remitting; SP = secondary
`progressive.
`
`—0.55% i 0.71%, SP = —0.63% i 0.69%,p =
`0.28) and so remained after correcting for age, dis-
`ease duration, and baseline NBV (p > 0.10 for all).
`This difference between the 2 datasets (trial/nontrial)
`
`was mainly due to the higher PBVC/y values found
`in trial than in nontrial data in both patients with
`C15 (PBVC/y, mean i SD: —0.49% t 0.50% vs
`—0.29% i 0.43%, p = 0.01) and RR patients
`(PBVC/y, mean : SD: —0.55% i 0.71% vs
`-0.34% 1' 0.45%, p < 0001) (figure 3). These
`differences persisted after controlling the PBCV/y
`values for age, disease duration, and baseline NBV
`(1) < 0.05 for all).
`Interestingly, most (64%) of the patients with
`C15 who converted to clinically definite MS were
`from trial data and showed higher PBVC/y than con-
`
`verted CIS from nontrial data (—0.58% i 0.55% vs
`
`—0.38% i 0.30%,p = 0.10).
`
`Relationships of PBVC/y with clinical-demographic
`features. Overall, the correlations of PBVC/y with
`
`both demographic and clinical features were weak or
`absent.
`
`Values of PBVC/y did not correlate with age, sex,
`or disease duration in the whole patient population,
`whereas they correlated weakly with baseline EDSS
`score (r = —0.15, p < 0.001) and AEDSS/y (r =
`—0.10,p = 0.003).
`In the different MS subtypes, a weak correlation was
`found in C15 between PBVC/y and age (r = 0.20, p =
`0.01) anddismseduration (r = 0.18,p = 0.02). In RR,
`PBVC/y correlated with age (r = 0.10, p = 0.04) and
`
`Neurology 74 June 8, 2010
`1871
`Page 4 of 9
`Copyright © by AAN Enterprises. Inc. Unauthorized reproduction of this article is prohibited.
`
`

`

`Figure 2
`
`
`
`Plot Illustrating overlap of baseline normalized brain volume (NBV)
`values for patients with clinically Isolated syndrome (CIS) and
`secondary progressive (SP) patients
`
`
`
`PBVC/y(%)
`
`
`
`1000
`
`1100
`
`1200
`
`1300
`
`See Results for details. PBVC = percent brain volume change.
`
`NBV (mm3)
`
`baseline EDSS score (r = —0.12,p = 0.003). No sig—
`nificant correlations were found in SP and PP.
`
`When data were grouped for source type, both
`populations of trial and nontrial data showed weak
`correlations between PBVC/y and baseline EDSS
`score (r = —0.21 and r = —0.12, p < 0.005 for
`both). All the other correlations were either weaker
`or absent (data not shown).
`
`Sample size estimation for PBVC/y. The sample sizes
`required to give a statistical power of 90% at a signif-
`icance level of 5% for different treatment effects
`
`(30%, 50%, and 70%) are summarized in table 2.
`
`Since there were differences in clinical-demographic
`characteristics (table e—l on the Neurology® Web site
`at www.neurology.org) as well as in PBVC/y values
`across disease subtypes when data were grouped for
`data source (trial vs nontrial), the sample sizes were
`estimated in these 2 dilferent datasets. Lower sample
`size estimation was found for patients with C15 in
`trial (n = 70) than in nontrial (n = 170) data. By
`contrast, no differences were found for R patients
`and SP patients.
`
`DISCUSSION While measures of brain atrophy
`rates have been used as an endpoint in many MS
`treatment trials, very little is known about the com—
`
`plex structural and temporal mechanisms leading to
`brain atrophy in the different MS subtypes. In panic-
`ular, the estimates of brain atrophy rates have varied
`in studies directly comparing different MS subtypes.
`These have shown higherls'26 or similar2223 atrophy
`progression in SP patients when compared to those
`at earlier disease stages. Moreover, atrophy rate in
`different MS subtypes showed high variability in the
`placebo arms of clinical
`trials. Thus, we collected
`here a large number of longitudinal MRI data with a
`median follow—up of 14 months from untreated pa—
`tients with MS with different disease subtypes and
`tested for differences in annualized brain atrophy
`rates in such a large population. We found heteroge-
`neity in PBVC/y across MS subtypes, which was par-
`ticularly evident when comparing patients at earliest
`with those at later disease stages. Interestingly, how-
`ever, this heterogeneity disappeared when PBVC/y
`values were corrected for the baseline NBV. This
`
`suggests that, nulling out the differences in atrophy
`state, atrophy progression rate is very similar in the
`different MS subtypes and, at late disease stages, does
`not seem to show either the previously hypothesized
`nonlinear progressionmz or a true acceleration.25'2‘
`This hypothesis is further supported by the finding
`that the absolute (in cm3) unadjusted changes were
`similar across disease subtypes when estimated, at the
`first order of approximation, by using the brain vol—
`ume at the 2 time points as recorded in the so-called
`halfway space of the SIENA method32 (data not
`shown).
`Since it is well—known that measures of brain vol-
`
`ume changes ean be significantly influenced by treat—
`ment with DMAR’3 we have included in this study
`only patients who, during the entire follow—up pe—
`riod, were either in the placebo arms of clinical trials
`or untreated in natural history studies. Interestingly,
`in the analysis of these patient populations, we found
`that annualized brain atrophy rates were signifieantly
`higher in trial than in nontrial data in both patients
`with C15 and RR patients. Thus, these data provide
`direct evidence on how differences in patients’ re—
`cruitment could significantly influence a measure
`such as brain atrophy rate even in a presumably ho-
`mogeneous patient population. Assuming that pa-
`tients with MS enrolled in a clinical trial are clinically
`more active than patients who decided to remain un—
`treated, these results also suggest that significant dif—
`ferences in brain atrophy progression may exist
`between populations of untreated patients with a
`similar MS subtype and different disease severity.
`This seems particularly true in patients with C15
`
`converting to clinically definite MS, who showed an
`almost twice as high brain atrophy rate in trial than
`in nontrial data.
`
`1872 Page 5 of 9
`Copyright © by AAN Enterprises. Inc. Unauthorized reproduction of this article is prohibited.
`
`Neurology 74 June 8, 2010
`
`

`

`Figure 3
`
`Percent braln volume change (PBVC)Iy values from nontrlal and trlal data In the different multiple
`
`sclerosis (MS) subtypes
`
`0.4
`
`0.2
`
`0.0
`
`PBVC/ycm a .b
`
`
`
`I Nontrlal data
`
`_1_4
`
`p<0.001
`
`N.S.
`
`ITrial data
`
`Columns and error bars represent means and standard deviations of PBVC/y. Note the significantly higher PBVC/y in
`patients with clinically isolated syndrome (CIS) and relapsing-remitting (RR) patients of the trial data MS group. See Results
`for details P = primary progressive; SP = secondary progressive.
`
`Many previous studies have investigated the
`relationship between brain atrophy and clinical
`disability.‘-35 As mentioned before for the patients
`with C15 and previously demonstrated in many stud—
`ies on different MS stages, brain atrophy rates are
`greater in patients who are worsening clinically than
`
`Table 2
`
`power. 5% significance level)“
`
`Sample size estimates to demonstrate
`a reduction in PBVC/y in patients
`with different MS subtypes in
`placebo-controlled trials (90%
`
`MS subtype and
`source
`CIS
`
`Nontrial
`
`Trial
`
`RR
`
`Nontn'al
`
`Trial
`
`SP
`
`Nontrial
`
`Trial
`
`PP
`
`Nontrial
`
`Treatment effect
`
`30%
`
`50%
`
`70%
`
`460
`
`180
`
`280
`
`240
`
`170
`
`140
`
`220
`
`170
`
`70
`
`110
`
`90
`
`64
`
`58
`
`81
`
`94
`
`40
`
`58
`
`54
`
`36
`
`32
`
`44
`
`Abbreviations: CIS = clinically isolated syndrome; M8 =
`multiple sclerosis; PBVC = percent brain volume change;
`PP = primary progressive; RR = relapsing—remitting; SP =
`secondary progressive.
`5 Estimates are the numbers of patients required for each
`trial arm to detect a treatment effect of 30%. 50%, and
`70%.
`
`in those who are not."25'35 While this suggests that
`measures of brain atrophy should have relevance on
`
`clinical progression, the correlation of global mea—
`sures of brain atrophy rate with clinical disability is
`usually far from being strict.”5 The results of the
`present study are consistent with this notion, show-
`ing that, even in a very large population spanning the
`whole spectrum of MS stages and severity, the corre-
`lation between clinical and atrophy measures tends
`
`to be weak. Perhaps selective measures of gray matter
`volume loss could improve this correlation, especially
`when combined with measures of cognitive impair-
`ment57 or related to disability status in the long
`term.” However, it is also true that in a complex
`disease such as MS several other factors (e.g., spinal
`cord damage, cortical adaptation to injury, genetic
`
`predisposition to injury repair, limits of the EDSS
`score to show disease progression) can all contrib-
`ute to weaken this relationship and should be con-
`sidered when interpreting the clinical relevance of
`this measure.
`
`As brain atrophy estimation has already been used
`as an outcome measure in numerous MS clinical tri-
`
`als, it is important to establish its sufficient power to
`detect a treatment effect. Thus, we estimated in all
`
`MS subtypes the sample size required to demonstrate
`a reduction of brain atrophy in patients in placebo—
`controlled trials. This was never done before for pa-
`tients with CIS and PP patients and was assessed
`recently in a small number of untreated RR patients
`(n = 33059 and placebo—treated SP patients (n =
`43).'7 In these 2 studies, which also measured PBVC
`
`Page 6 of 9
`Copyright © by AAN Enterprises. Inc. Unauthorized reproduction of this article is prohibited.
`
`Neurology 74 June 8, 2010
`
`1873
`
`

`

`by using SIENA, the sample sizes required to have
`90% power and 50% treatment effect in a 1-year
`trial was very similar to our estimation in SP patients
`(63 vs 58, when comparing data from their placebo-
`treated patients with our trial data), but lower than
`our estimation in RR patients (69 vs 110, when com-
`paring data from their untreated patients with our
`nontrial data). The lower sample size estimated in the
`previous study for RR patients is consistent with the
`higher atrophy rate estimated over 1 year in that
`study39 compared to the atrophy rate we detected in
`our much larger cohort of RR patients. Regardless,
`our data overall show that the sample size needed to
`demonstrate a reduction of brain atrophy in patients
`in a placebo-controlled trial is larger in patients with
`CIS and RR patients than in patients with progres-
`sive MS forms. It also seems much lower in the MS
`population included in treatment trials than in pa-
`tients untreated in clinical practice. It should be
`noted that the reported numbers would be higher if
`the physiologic atrophy rate detectable in healthy
`controls (about 0.1% in a previous study39) and the
`percent of unusable scan pairs (12% in the present
`study, therefore dividing the resulting sample sizes by
`0.882) were taken into account.
`There are some limitations to our study. First, the
`analysis of brain atrophy rates was applied retrospec-
`tively to existing T1-weighted MRI datasets. These
`included MRIs that were obtained with different ac-
`quisition protocols, prior to or after injection of gad-
`olinium and with different slice thicknesses. All these
`variables have been taken into account in the statisti-
`cal analysis and previous studies have shown that the
`measurement obtained with SIENA is relatively in-
`sensitive to them.2,32,34 It must be stressed, however,
`that the SIENA method, similarly to other registration-
`based methods,2 requires the pair of images of the
`same subject to have identical acquisition parameters
`and an adequate repositioning of the follow-up scans.
`These important requirements were ensured here
`and this explains the relatively high number (12%) of
`MR examinations that were excluded from the anal-
`ysis. Second, we provided global measures of PBVC,
`without performing a tissue-type (e.g., white matter
`lesions, normal-appearing white and gray matter)
`analysis. Several segmentation algorithms are able to
`provide accurate measures of white and gray matter
`volumes. However, they are heavily dependent on
`image characteristics (e.g., intensity, contrast, and
`slice thickness), which were very heterogeneous in
`our study cohort. By contrast, while the SIENA ap-
`proach (by assessing edge displacements between 2
`brains) is quite insensitive to image intensity varia-
`tions, it does not allow white/gray matter segmenta-
`tion.32 Finally, the length of the study follow-up was
`
`relatively short (median of 14 months) and did not
`vary in a similar magnitude across MS subtypes. The
`use of annualized measures, however, should have
`corrected for this.
`Despite limitations, results of this study suggest
`that brain atrophy proceeds relentlessly throughout
`the course of MS, with a rate that seems largely inde-
`pendent of the MS subtype. There is not a simple
`relationship between clinical disability and atrophy
`progression, at least in the short term. Moreover, the
`rate of atrophy significantly differed between the MS
`population recruited for trials and those who re-
`mained untreated in a clinical standard setting. Thus,
`brain atrophy rates do have some clinical relevance
`and heterogeneity underlying patients with MS even
`in the same subtypes could be, to some extent, dis-
`closed by measuring the atrophy progression rate.
`
`AUTHOR CONTRIBUTIONS
`N.D.S. coordinated the study and wrote the manuscript, with contribu-
`tions from A.G., M.R., M.P.S., F.B., C.E., F.F., A.T., and M.F. M.B. and
`A.G. were responsible for MRI data analysis. A.G. and M.P.S. did statis-
`tical analysis. A.G. was responsible for data collection. M.R., F.B., T.K.,
`C.E., F.F., M.C., D.D., G.T., A.G., X.M., A.R., A.T., G.C., D.H.M.,
`and M.F. provided MRI data.
`
`DISCLOSURE
`Prof. De Stefano has served on scientific advisory boards for Teva Phar-
`maceutical Industries Ltd., BioMS Medical, Biogen-Dompe´ AG, and
`Merck Serono; has received funding for travel from Teva Pharmaceutical
`Industries Ltd. and Merck Serono; has received speaker honoraria from
`Teva Pharmaceutical Industries Ltd., BioMS Medical, Biogen-Dompe´
`AG, Merck Serono, and Bayer Schering Pharma; and has received research
`support from the Italian MS Society. Dr. Giorgio and Mr. Battaglini
`report no disclosures. Dr. Rovaris has received funding for travel from
`Teva Pharmaceutical Industries Ltd. and Biogen-Dompe´ AG; has received
`speaker honoraria from Teva Pharmaceutical Industries Ltd., Biogen-
`Dompe´ AG, sanofi-aventis, and Bayer Schering Pharma; and has received
`research support from the Italian MS Society. Dr. Sormani has received
`speaker honoraria from Biogen Idec, Merck Serono, Teva Pharmaceutical
`Industries Ltd., and Biogen-Dompe´ AG; and serves as a consultant for
`Merck Serono, Actelion Pharmaceuticals Ltd, Eidetica, and Biogen Idec.
`Dr. Barkhof serves on scientific advisory boards for Lundbeck Inc.,
`Roche, Bayer Schering Pharma, sanofi-aventis, UCB, and Novartis; serves
`on the editorial boards of Brain, the Journal of Neurology, Neurosurgery,
`and Psychiatry, European Radiology, the Journal of Neurology, and Neurora-
`diology; and has received compensation for consultancy, committee mem-
`bership and lecture fees from Bayer Schering Pharma, Merck Serono SA,
`sanofi-aventis, AstraZeneca, Genentech, Inc., Novartis, Biogen Idec, Ac-
`tive Biotech, European Charcot Foundation, Lundbeck Inc., Talecris Bio-
`therapeutics, Roche, UCB, Wyeth, MediciNova, Inc. Dr. Korteweg and
`Dr. Enzinger report no disclosures. Prof. Fazekas has served on scientific
`advisory boards for Biogen Idec, Teva Pharmaceutical Industries Ltd., and
`sanofi-aventis; serves on editorial advisory boards for Cerebrovascular Dis-
`eases, the Journal of Neurology, the Polish Journal of Neurology and Neuro-
`surgery, Stroke, and Swiss Archives of Neurology and Psychiatry; has received
`honoraria from Merck Serono and Bayer Schering Pharma; and has re-
`ceived research support from Biogen Idec, Merck Serono, Teva Pharma-
`ceutical Industries Ltd., and sanofi-aventis. Dr. Calabrese and Dr. Dinacci
`report no disclosures. Prof. Tedeschi serves on the editorial board of Neu-
`rological Sciences; has received speaker honoraria from Merck Serono,
`Bayer Schering Pharma, Teva Pharmaceutical Industries Ltd., sanofi-
`aventis, and Biogen-Dompe´ AG; and receives research support from Sec-
`ond University of Naples and from the Italian Ministry of Health. Dr.
`Gass serves on the editorial board of Cerebrovascular Diseases; and has
`
`1874
`
`Neurology 74 June 8, 2010
`
`Page 7 of 9
`
`

`

`received speaker honoraria from Bayer Schering Pharma, Biogen Idec,
`Teva Pharmaceutical Industries Ltd. and Merck Serono. Dr. Montalban
`serves on scientific advisory boards for Novartis, Teva Pharmaceutical
`Industries Ltd., Merck Serono, Biogen Idec, and Bayer Schering Pharma;
`has received funding for travel and speaker honoraria from Novartis, Teva
`Pharmaceutical Industries Ltd., Merck Serono, Biogen Idec, sanofi-
`aventis, and Bayer Schering Pharma; serves on the editori